This proposal investigates a novel ankle robot (anklebot) adaptive control approach integrated with treadmill training to reduce foot drop and improve mobility function in chronic hemiparetic stroke survivors. Currently, stroke survivors with foot drop are trained to live with a cane or othr assistive device, and often ankle foot orthotics (AFO's) for safety. Neither mediates task-practice or neuromotor recovery. We have developed an adaptive anklebot controller that detects gait cycle sub-events for precise timing of graded robotics assist to enable deficit severity adjusted ankle motor learning in the context of walking. Our Merit pilot findings show that 6 weeks treadmill training with robot (TMR) timed to assist swing phase dorsiflexion only, is more effective than TM alone to improve free-walking swing dorsiflexion at foot strike (+400%, +9.7, N=4/group), floor-walking speed (21% vs. 9%), and the benefits are retained at 6 weeks post- training such that 3 of 4 participants no longer required their AFO's. Notably, swing-phase TMR training improved paretic leg push-off (277% vs. 2%), and center-of-pressure (CoP) sway on standing balance (-25% vs. +30%), indicating generalized benefits to other elements of gait and balance, beyond those robotically targeted toward foot drop. This randomized study investigates the hypothesis that 6 weeks TMR is more effective to durably improve gait biomechanics, static and dynamic balance, and mobility function in chronic stroke survivors with dorsiflexion deficits, compared to TM alone. Aims are to determine the compare effectiveness of 6 weeks TMR vs. TM alone on: 1) Independent gait function indexed by gait velocity, swing-phase DF, terminal stance push off, and 48- hour free-living mobility profiles. 2) Balance function indexed by measures of postural sway (CoP), asymmetric loading in quiet standing, peak paretic A-P forces in non-paretic gait initiation, and standardized scales (Berg Balance, Dynamic Gait Index, and ABC for Falls Efficacy and; 3) Long-term mobility outcomes, assessed by repeated measures of all key gait and balance outcomes at 6 weeks and 6 months after formal training cessation. While upper extremity robotics has proven effective, and altered national care standards for stroke, in contrast, lower extremity robotics that has primaril focused on repetitive, multi-joint patterning of gait cycles, remains controversial, with consensus that current approaches are inferior to usual care, or even deleterious. We challenge this paradigm by testing an adaptively controlled anklebot guided by sensorimotor learning models to focus specifically on complications due to impaired paretic DF control. The profile of ankle motor learning based on repeat measures of unassisted walking reveals a power law pattern with ~80% of gains <3 weeks. Hence, TMR may prove effective within the time-frame of usual physical therapy, which increases potential for translation into VA care. Results of this study wil establish a new paradigm for diverse deficit severity customized gait-integrated adaptive modular LE robotics to improve mobility function in stroke and other neurological conditions.